It will be observed that the composite of 56 female faces is made by
the blending of two other composites, both of which are given. The
history was this--I took the 56 portraits and sorted them into two
groups; in the first of these were 20 portraits that showed a
tendency to thin features, in the other group there were 36 that
showed a tendency to thickened features. I made composites of each
of them as shown in the Plate. Now it will be remarked that,
notwithstanding the attempt to make two contrasted groups, the
number of mediocre cases was so great that the composities of the
two groups are much alike. If I had divided the 56 into two
haphazard groups, the results would have been closely alike, as I
know from abundant experience of the kind. The co-composite of the
two will be observed to have an intermediate expression. The test
and measure of statistical truth lies in the degree of accordance
between results obtained from different batches of instances of the
same generic class. It will be gathered from these instances that
composite portraiture may attain statistical constancy, within
limits not easily distinguished by the eye, after some 30 haphazard
portraits of the same class have been combined. This at least has
been my experience thus far.

The two faces illustrative of the same type of tubercular disease
are very striking; the uppermost is photographically interesting as
a case of predominance of one peculiarity, happily of no harm to the
effect of the ideal wan face. It is that one of the patients had a
sharply-checked black and white scarf, whose pattern has asserted
itself unduly in the composite. In such cases I ought to throw the
too clearly defined picture a little out of focus. The way in which
the varying brightness of different pictures is reduced to a uniform